A Top Down Approach to Enterprise Monitoring Using Change Point Detection

摘要

Monitoring is a key aspect of IT service management of any enterprise. Besides providing visibility into the IT infrastructure, it provides information required to manage resources, plan for growth and ensure uptime and availability of applications. At the higher levels of monitoring that are vital to service management, key business and IT processes are constantly monitored to collect valuable data about the performance and availability of the processes. Traditionally, in the bottom-up approach towards monitoring, these IT and business metrics may be defined as an aggregation of multiple lower level metrics which in turn may be defined as an aggregation of other metrics. The complexity of these business and IT processes is captured through modeling (e.g.: CIM) which is both time consuming and requires a high level of skill and understanding of the infrastructure. It requires an in-depth understanding of the various systems that work together to provide the services that need to be monitored. In large enterprises where business and IT change rapidly to adapt to market requirements, configuring a monitoring system to monitor new processes by creating accurate models can be a challenge. Due to the time required to construct these monitoring models, there is a temporal lack of visibility into the behavior of the processes which is a matter of concern to the business and IT teams in the enterprise. In this paper, we present a top-down approach towards monitoring high level business and IT processes. We present a framework that monitors the highest level metrics and detects changes in their behavior without the need for comprehensive modeling, thereby allowing these processes to be quickly and efficiently monitored. We also present examples and results from a deployment of the system in the IT environment of a large telecom company.